Professional Documents
Culture Documents
Efficacy of Helicobacter Pylori Eradication Therapy For Functional
Efficacy of Helicobacter Pylori Eradication Therapy For Functional
Original research
Figure 1 Forest plot of RCTs of H. pylori eradication therapy: effect on symptom cure in FD. FD, functional dyspepsia; H. pylori, Helicobacter pylori;
RCTs, randomised controlled trials.
Ford AC, et al. Gut 2022;71:1967–1975. doi:10.1136/gutjnl-2021-326583 1969
Helicobacter pylori
Figure 2 Forest plot of RCTs of H. pylori eradication therapy: effect on symptom improvement in FD. FD, functional dyspepsia; H. pylori, Helicobacter
pylori; RCTs, randomised controlled trials.
antibiotics, or placebo alone, the RR of symptoms not improving 16). Limiting the analysis to only the six RCTs at low risk of bias
was 0.91 (95% CI 0.89 to 0.94), with an NNT of 14 (95% still demonstrated a significant effect of H. pylori eradication
CI 12 to 22) with no heterogeneity between studies (I2=0%, therapy (RR of symptoms not being cured or improving=0.90;
p=0.80), and no evidence of funnel plot asymmetry (Egger test, 95% CI 0.81 to 0.99, I2=46%, p=0.10) (online supplemental
p=0.24). When the eight RCTs that compared antisecretory figure 3),20 43 49 55 58 62 with an NNT of 14 (95% CI 7 to 139).
therapy and placebo antibiotics with eradication therapy were When we performed a subgroup analysis based on study loca-
included,17 42 43 45 47 48 58 59 there was still a benefit (RR=0.92; tion, the effect of H. pylori eradication therapy was greater in
95% CI 0.88 to 0.96, I2=0%, p=0.73), with an NNT of 16 (95% Asian studies (RR=0.82; 95% CI 0.73 to 0.91, NNT=8; 95% CI
CI 11 to 33). Similarly, when only the six trials that compared 5 to 15, I2=77%, p<0.001) compared with non-Asian studies
placebo alone with eradication therapy were included,18–20 50 55 57 (RR=0.92; 95% CI 0.89 to 0.95, NNT=16; 95% CI 12 to 25,
the benefit remained (RR=0.91; 95% CI 0.87 to 0.96, I2=0%, I2=11%, p=0.32) and this difference was statistically significant
p=0.47) with an NNT of 14 (95% CI 10 to 31). (p value for χ2=0.04) (online supplemental figure 4). Finally,
There were 22 trials that reported on improvement of FD when we performed a subgroup analysis based on study dura-
symptoms,18–20 38–41 44–47 49–54 57 58 60–62 randomising 5193 H. tion, a beneficial effect on FD symptoms was seen in studies of
pylori-positive patients with FD, 2824 of whom received erad- 12-month duration (RR=0.87; 95% CI 0.82 to 0.92, NNT=10;
ication therapy. The RR of FD symptoms not improving with 95% CI 7 to 17, I2=65%, p<0.001), but not studies of <12-
eradication therapy versus control was 0.84 (95% CI 0.78 month duration (RR=0.88; 95% CI 0.77 to 1.00, I2=67%,
to 0.91) (figure 2), with moderate heterogeneity between p=0.01) (online supplemental figure 5), but this difference was
studies (I2=69%, p<0.001). The funnel plot was symmetrical, not statistically significant (p value for χ2=0.87).
suggesting no evidence of publication bias or other small study
effects (Egger test, p=0.16). The NNT was 9 (95% CI 7 to 17).
When we limited the analysis to only 16 studies, containing Effect of H. pylori eradication therapy on FD symptoms
4224 patients,18–20 38 39 41 44–47 49 50 57 58 61 62 that compared erad- according to eradication rates, success or failure of
ication therapy with either antisecretory therapy with placebo eradication therapy, and antibiotics used
antibiotics or placebo alone, the RR of symptoms not improving To assess whether the beneficial effect of eradication therapy
was 0.86 (95% CI 0.79 to 0.94), with an NNT of 10 (95% CI was due to eradication of H. pylori or an effect related to the
7 to 24). Moderate heterogeneity between studies persisted use of antibiotics in general, or a particular antibiotic, we first
(I2=71%, p<0.001), but there was no evidence of funnel plot assessed for any correlation between eradication rates observed
asymmetry (Egger test, p=0.46). Eight of these trials compared in the trials and rates of symptom cure or improvement using a
antisecretory therapy and placebo antibiotics with eradication Pearson correlation coefficient. There was no significant correla-
therapy (RR=0.83; 95% CI 0.73 to 0.95, I2=81%, p<0.001; tion between the two (Pearson correlation coefficient=−0.23,
NNT=8; 95% CI 5 to 27),38 39 44–47 58 61 and eight compared p=0.907) (figure 3).
placebo alone with eradication therapy (RR=0.90; 95% CI 0.80 Second, we compared rates of symptom cure or improve-
to 1.00, I2=49%, p=0.06).18–20 41 49 50 57 62 ment according to whether H. pylori was eradicated success-
Pooling data from all 29 RCTs, irrespective of whether they fully in patients in the trial arms that received active H.
reported symptom cure or improvement, but using symptom pylori eradication therapy. In these analyses, there were 16
cure preferentially, where reported, the RR of symptoms not RCTs,18 19 38 40 41 43 48–50 54–57 59–61 containing 2809 patients,
being cured or not improving with H. pylori eradication therapy where symptom cure or improvement rates could be compared
versus control was 0.87 (95% CI 0.83 to 0.92), with moderate between patients with successful eradication of H. pylori and
heterogeneity between studies (I2=64%, p<0.001) (online patients in the control arms. The RR of symptoms not being
supplemental figure 2). Again, the funnel plot was asymmetrical, cured or not improving with successful H. pylori eradication
suggesting evidence of publication bias or other small study therapy versus control therapy was 0.74 (95% CI 0.64 to 0.85,
effects (Egger test, p=0.014). The NNT was 10 (95% CI 8 to I2=82%, p<0.001) (figure 4), and with evidence of funnel plot
1970 Ford AC, et al. Gut 2022;71:1967–1975. doi:10.1136/gutjnl-2021-326583
Helicobacter pylori
Figure 3 Scatterplot of H. pylori eradication rate versus rate of symptom cure or improvement in individual RCTs. H. pylori, Helicobacter pylori; RCTs,
randomised controlled trials.
asymmetry (Egger test, p=0.076). The NNT was 6 (95% CI 4 amoxicillin and a nitroimidazole, or clarithromycin and a
to 10). Thirteen trials,19 38 40 41 43 48 49 55–57 59–61 containing 1517 nitroimidazole.
patients, provided data comparing symptom cure or improve- Given this, we assessed for differences in efficacy across
ment rates between patients with unsuccessful eradication of H. all nitroimidazole-containing and non- containing regimens,
pylori and patients in the control arms. In this analysis, the RR as well as nitroimidazole-containing and non- containing PPI
of symptoms not being cured or not improving with unsuccessful triple therapy regimens (online supplemental figures 7 and 8).
H. pylori eradication therapy versus control therapy was 1.02 However, there were no significant differences between these
(95% CI 0.96 to 1.09, I2=0%, p=0.63) (online supplemental subgroup analyses according to the χ2 test (p=0.10 and p=0.17,
figure 6), with no evidence of funnel plot asymmetry (Egger respectively).
test, p=0.75). These same 13 RCTs,19 38 40 41 43 48 49 55–57 59–61
containing data from 1259 patients, compared symptom cure
or improvement rates between patients receiving eradication Adverse events
therapy with successful or unsuccessful eradication of H. pylori. Adverse events data were reported incompletely by most trials.
The RR of symptoms not being cured or not improving with Only eight RCTs reported total numbers of adverse events, in
successful versus unsuccessful eradication of H. pylori was 0.65 1937 patients.19 41 45 46 48 52 56 58 Overall, adverse events were
(95% CI 0.52 to 0.82, I2=79%, p<0.001) (figure 5), with no significantly more common with eradication therapy (RR=2.19;
evidence of funnel plot asymmetry (Egger test, p=0.20). The 95% CI 1.10 to 4.37), with high levels of heterogeneity between
NNT in this analysis was 4.5 (95% CI 3 to 9). studies ((I2=92%, p<0.001) (online supplemental figure 9)
Finally, we compared efficacy of eradication therapy with and an NNH of 3 (95% CI 1 to 40). Adverse events leading
control therapy according to the eradication regimen or anti- to withdrawal were reported by 18 trials, recruiting 3694
biotics used (table 1). NNTs were lower, generally, with patients.18–20 38 39 41–43 45–47 49 50 52 53 55–57 Withdrawals due to
nitroimidazole-containing regimens, and in regimens consisting adverse events were also significantly more likely with erad-
of a proton pump inhibitor (PPI) in combination with either ication therapy (RR=2.60; 95% CI 1.47 to 4.58), with no
Figure 4 Forest plot of RCTs of H. pylori eradication therapy: effect on symptom cure or improvement in FD in those with successful H. pylori
eradication therapy versus control therapy. FD, functional dyspepsia; H. pylori, Helicobacter pylori; RCTs, randomised controlled trials.
Ford AC, et al. Gut 2022;71:1967–1975. doi:10.1136/gutjnl-2021-326583 1971
Helicobacter pylori
Figure 5 Forest plot of RCTs of H. pylori eradication therapy: effect on symptom cure or improvement in FD in those with successful H. pylori
eradication therapy versus unsuccessful H. pylori eradication therapy. FD, functional dyspepsia; H. pylori, Helicobacter pylori; RCTs, randomised
controlled trials.
heterogeneity between studies (I2=0%, p=0.66) (online supple- pylori eradication rates and rates of symptom cure or improve-
mental figure 10) and an NNH of 71 (95% CI 32 to 242). There ment, which suggests the effect may relate to a general impact
was no evidence of funnel plot asymmetry (Egger test, p=0.62). on the gut microbiota. This analysis has the advantage that all
data are evaluated and so publication bias is less likely, but as
DISCUSSION trial level data are being used it is possible that the results are
This updated systematic review and meta-analysis has confirmed confounded by trial level factors. On the other hand, the RR
that eradication therapy is efficacious for the treatment of FD of FD symptoms not being cured or not improving was signifi-
in H. pylori-positive patients. The RR of either symptoms not cantly lower in those patients receiving eradication therapy who
being cured or symptoms not improving was significantly lower had successful eradication of H. pylori, compared with either
than with a control intervention, with an NNT of 14 and 9, patients receiving a control intervention or patients randomised
respectively. Adverse events were more common with eradi- to eradication therapy in whom eradication was unsuccessful,
cation therapy, as were adverse events leading to withdrawal, with an NNT of 6 and 4.5, respectively, which appear more
although reporting of these data was incomplete in many trials.
impressive. This suggests a specific effect of the eradication
Although the effect on cure of symptoms was modest, when
regimen on H. pylori, although over half the trials did not
applied to entire healthcare systems, and for a highly preva-
report these data, so publication bias is probable. Also, patients
lent condition, this is likely to lead to substantial reductions in
management costs. Eradication therapy is particularly likely to remaining H. pylori-positive are likely to be different to those
be cost effective as a 2-week course of treatment has an impact with successful eradication, and so these results may also be due
for at least 12 months. We judged the evidence for efficacy and to bias or confounding. For instance, patients in whom eradica-
safety of eradication therapy in H. pylori-positive patients with tion of H. pylori was unsuccessful may be less likely to adhere to
FD as high quality (online supplemental table 4). the trial medication, and hence did not receive antibiotics at all.
Whether this relates to the eradication of H. pylori or a more Finally, NNTs were lower with H. pylori eradication regimens
general impact on the upper gastrointestinal tract microbiota is containing a nitroimidazole, although these differences in effi-
unclear from our data. There was no correlation between H. cacy were not statistically significant.
Table 1 Relative risk of symptoms not being cured or not improving and NNT according to H. pylori eradication regimen used
RR of symptoms not being
Number of cured or not improving NNT P value
H. pylori eradication regimen used Number of trials patients (95% CI) (95% CI) I2 for χ2
Amoxicillin containing 25 6076 0.91 15 19% 0.20
(0.88 to 0.94) (11 to 22)
Clarithromycin containing 22 5577 0.87 10 67% <0.001
(0.82 to 0.92) (7 to 17)
Nitroimidazole containing 8 1589 0.79 6.5 86% <0.001
(0.67 to 0.93) (4 to 18.5)
Tetracycline containing 2 223 0.54 N/A 88% 0.005
(0.16 to 1.77)
PPI, amoxicillin and clarithromycin 17 4159 0.92 16.5 0% 0.61
(0.89 to 0.95) (12 to 26.5)
PPI, amoxicillin and nitroimidazole 2 469 0.87 8.5 0% 0.33
(0.81 to 0.94) (6 to 18)
PPI, clarithromycin and nitroimidazole 2 482 0.68 4 89% 0.003
(0.47 to 0.98) (2 to 57)
PPI triple therapy containing a 5 1112 0.80 6 86% <0.001
nitroimidazole (0.66 to 0.97) (4 to 40)
N/A, not applicable, as no significant benefit of treatment ; NNT, number needed to treat; PPI, proton pump inhibitor.